Environmental Impact

Jon Ippolito, Professor at the University of Maine, has developed an app called What Uses More to explore the environmental footprint of various digital activities (including AI prompts but also other practices, e.g. watching Netflix). His app encourages users to consider and understand electrical and water usage required by AI

Making Sense of Sustainability Reports. Peyton McCauley and Melissa Scanlan(University of Wisconsin-Milwaukee) discuss how data center companies don’t always reveal how much water their data centers use. 

The frontier model developers attempt to address the environmental impact that has attended the development of GenAI:

Microsoft’s 2026 Environmental Sustainability Report provides self-reporting details on their efforts towards environmental sustainability.

Google’s 2026 Environmental Report provides self-reporting details on their commitment to environmental sustainability.

Open Source LLM’s

Open weight and open source LLM’s are not the same thing. The distinction is important as Gary Marcus has observed on Substack (10 August 2026): “Open-weight piggybacks onto all the great press that open-source has but offers few of the advantages. Open-weight models are NOT fully customizable, and not transparent. They are the product of machine learning (the weights of a trained neural network), but not the whole thing. . . . That means you can run them (or “posttrain” them), but, crucially, users can’t customize or improve them.”

AI2

The Allen Institute for AI (AI2) has released fully open-source large language models — including Olmo 2 and Olmo 3 — designed to help researchers better understand how AI works. Among other possibilities, Open Source LLM’s allow academic and researchers one way to experiment and engage the issue “interpretability” (or the failure to understand how AI works).

Ollama

Ollama is a freely available, open source application that allows you to run LLM’s locally on your computer for experimentation and testing.